Veronika I. Korshunova

this study investigates the application of decision trees for detecting phishing domains based on URL attributes. A logical classification algorithm searching for conjunctive patterns was applied to a dataset comprising 112 structural and server-related features. The decision tree model achieved an error rate of 3.26%. Key significant factors identified include the quantity of slashes in the URL (40.3%), domain activation time (16.8%), and directory length (14.06%). The results confirm the high efficiency and interpretability of rule-based machine learning for cybersecurity threat mitigation.